Submissions

Author Guidelines

MLCS publishes research in which machine learning and a substantive application domain advance together. Before preparing a manuscript, please read the Focus and Scope page. It defines the four convergence conditions that editors use to assess fit, names the journal's principal domain, and lists what falls outside the scope. Manuscripts that do not fit the scope are declined at initial screening without external review.

1. Before you submit

Originality. Submissions must report original work that has not been published elsewhere, in any language, and is not under consideration by another journal. A manuscript that substantially reproduces an article already published, including one by the same authors, is not considered.

Preprints. Posting a preprint on a recognized preprint server, such as arXiv or medRxiv, is permitted and is not treated as prior publication. Disclose the preprint in the cover letter, and link it to the published article after acceptance.

Conference papers. An extended version of a conference paper is considered only if it contains substantial new work, cites the conference paper, discloses it in the cover letter, and is permitted by the conference publisher's policy.

Language. Manuscripts must be written in clear American English. A manuscript that requires substantial language revision may be returned to the authors before review.

Fees. MLCS does not charge submission or article processing fees.

2. Article types

Research article (up to 8,000 words), original work that meets the four convergence conditions. Review article (up to 10,000 words), a systematic or scoping review that follows PRISMA 2020 or PRISMA-ScR, with the completed checklist, a flow diagram, and, where applicable, a registered protocol; or a review invited by the editors. Unsolicited narrative reviews are not considered.

3. Preparing your manuscript

Format at first submission. The journal template is not required at first submission. Submit a single-column manuscript in DOCX, or as a PDF compiled from LaTeX, with numbered pages and continuous line numbers. Place figures and tables near their first mention or at the end of the file. The MLCS template is required for the revised manuscript once the editors request minor revisions or accept the manuscript.

Double-anonymous review. MLCS uses double-anonymous peer review. Upload two separate files:

• a title page containing all identifying information, as listed below; and
• an anonymized manuscript containing no author names, affiliations, acknowledgments, funding details, or other identifying information, including in the file properties.

In the anonymized manuscript, cite your own earlier work in the third person (for example, "Smith et al. [5] showed" rather than "our previous work [5] showed"). Replace links to code or data repositories that identify you with anonymized links, or provide the materials to the editors as confidential supplementary files.

Title page. The title page must include:

• the title;
• the full name of every author, with department, institution, city, and country;
• the ORCID iD of every author (required for the corresponding author and strongly encouraged for all);
• the corresponding author's institutional email address;
• the CRediT author contribution statement (Section 7);
• the funding statement and the declaration of competing interests (Section 9);
• acknowledgments; and
• the statement on the use of generative AI (Section 8).

Structure of a research article.

• Title: specific and informative, without unexplained abbreviations.
• Abstract: a single paragraph of no more than 250 words, without citations or undefined abbreviations. It should state the context, the gap or objective, the methods, the main results with key quantitative findings, and the conclusion.
• Keywords: four to six.
• Introduction: the problem, why it matters in the domain, what is already known, and the specific contribution of the study.
• Methods: enough detail for an independent group to reproduce the study (Section 4).
• Results: findings reported with measures of uncertainty.
• Discussion: interpretation, comparison with prior work, limitations, and implications for the domain.
• Conclusion.
• Back matter and references.

Section headings may be adapted to the study, but every element above must be present.

4. Reporting standards

MLCS expects authors to provide the evidence a reader needs to judge whether a result holds and whether it matters in its domain. The points below are checked at initial screening and in peer review.

Data and evaluation design

• Describe the data source, inclusion criteria, sample size, class distribution, and preprocessing.
• State how the data were divided into training, validation, and test sets, and at what level. When several samples come from one person, split by person rather than by sample.
• Perform every data-dependent step, including normalization, imputation, feature selection, and resampling for class imbalance, within the training portion of each split only. Steps applied to the full dataset before splitting leak information into the evaluation and invalidate it.
• Where possible, validate on data from a different source, site, or period than the training data.

Baselines and tuning

• Compare against appropriate and current baselines, run under the same protocol, on the same splits, and with a comparable tuning budget. Results quoted from other publications may be reported for context, but they cannot support a claim of superiority.
• Describe the hyperparameter search space, the search procedure, and the data on which tuning was performed. Tuning on the test set is not acceptable.

Metrics and statistics

• Choose metrics suited to the decision at stake. For imbalanced problems, report metrics that reflect minority-class performance, such as sensitivity, specificity, balanced accuracy, G-mean, F1, or PR-AUC, and not accuracy alone.
• Report variability across repeated runs, folds, or bootstrap samples, not point estimates alone.
• When claiming that one method outperforms another, use and name an appropriate statistical test. Account for the dependence between cross-validation folds, and correct for multiple comparisons when several are made.
• When model outputs are used as probabilities, report calibration.

Contribution of components

• When a method combines several components, show the contribution of each, for example through an ablation study, so that readers can tell which component is responsible for the result.

Negative and null results

• Well-conducted studies with negative or null results are welcome, particularly as short communications.

Reporting guidelines

Studies in the health and clinical domain should follow the relevant guideline from the EQUATOR Network and include the completed checklist as a supplementary file:

• TRIPOD+AI for clinical prediction models;
• CLAIM for artificial intelligence in medical imaging;
• STARD 2015 for diagnostic accuracy studies;
• CONSORT-AI and SPIRIT-AI for clinical trials and trial protocols involving artificial intelligence;
• DECIDE-AI for early-stage clinical evaluation of decision support systems; and
• PRISMA 2020 or PRISMA-ScR for systematic and scoping reviews.

5. Data and code availability

Data. Every research article must include a Data Availability Statement explaining where and how the data can be accessed.

• Where possible, deposit the data in a public repository that assigns a persistent identifier.
• Where the data cannot be shared publicly for legal, ethical, or privacy reasons, such as identifiable patient data, explain the restriction, name the data custodian, and describe how qualified researchers can request access.
• For publicly available datasets, cite the source and state the license.
• A statement that data are "available on request" is not accepted without the reason for the restriction and the name of the custodian.

Code. For manuscripts that propose or evaluate a method, the code needed to reproduce the main results must be made available. During review, share it through an anonymized link or as a confidential supplementary file. After acceptance, deposit it in a public repository and archive the released version with a persistent identifier, for example through Zenodo. Where code cannot be shared, explain why in a Code Availability Statement; the editors will then consider whether the claims can be assessed without it.

6. Research ethics

• Research involving human participants, human data, or human tissue must have been approved by an appropriate ethics committee or institutional review board. State the name of the committee and the approval number, and confirm that the research complied with the Declaration of Helsinki.
• State whether informed consent was obtained, or why the requirement was waived.
• Studies that use only publicly available, fully de-identified data must state this and cite the data source. Formal approval may not be required, but the basis for that conclusion must be stated.
• Remove all information that could identify a patient from the text, figures, and supplementary files.

7. Authorship

• MLCS follows the ICMJE criteria for authorship. Everyone who meets the criteria must be listed as an author, and every listed author must meet them. Other contributors should be acknowledged.
• A CRediT author contribution statement is required. For studies in a convergence domain, it should identify the authors who provide the domain expertise on which the study's domain claims rest.
• All authors must have approved the submitted version and agreed to its submission. The corresponding author communicates with the journal on behalf of all authors.
• A change to the author list after submission requires a request that explains the change, with the written agreement of all authors, including any author being added or removed.
• Submitting work without the consent of all its authors, or submitting work of which the submitter is not an author, is misconduct and will be handled according to COPE guidance.

8. Use of generative artificial intelligence

• Generative AI tools cannot be listed as authors, because they cannot take responsibility for the work.
• Disclose any use of generative AI in preparing the manuscript, including the tool, its version, and the purpose, such as language editing, drafting text, or generating code, in a statement placed before the references. Basic spelling and grammar checking need not be disclosed.
• Where AI tools were used in the research itself, for example to generate or annotate data or to write analysis code, describe this in the Methods.
• Authors are fully responsible for the accuracy, originality, and integrity of all content, including content produced with AI assistance, and must verify every reference against its source.
• AI-generated or AI-altered images are not permitted, except where such images are the subject of the research and this is described in the Methods.

9. Competing interests and funding

• All authors must declare any financial or non-financial interest that could be perceived as influencing the work. If there is none, state: "The authors declare no competing interests."
• List all sources of funding, with grant numbers, and describe the role of each funder in the study, or state that the funder had no role.

10. References

• Use numbered citations in square brackets, numbered in order of first appearance. At first submission, any consistent numbered style is acceptable; the journal style is applied through the template.
• Include the DOI for every reference that has one.
• Cite primary sources wherever possible, and verify every reference against the original publication.
• Limit self-citation to work that is directly relevant.
• MLCS editors and reviewers do not ask authors to cite articles published in MLCS. Please report any such request to the editorial office.

11. Figures, tables, and supplementary material

• Supply raster images at a resolution of at least 300 dpi. Vector formats (PDF, EPS, or SVG) are preferred for charts and diagrams.
• Use color schemes that remain distinguishable for readers with color vision deficiency, and do not rely on color alone to convey meaning.
• Provide tables as editable text, not as images.
• Cite every figure and table in the text, and give each a caption that can be understood without reading the main text.
• Obtain written permission to reproduce any figure or table published elsewhere, and cite the source.
• Supplementary material, such as additional results, checklists, code, and data, is reviewed with the manuscript and published as submitted.

12. Mathematics, units, and abbreviations

• Number displayed equations consecutively, and define every symbol at first use.
• Use SI units.
• Define each abbreviation at first use in the abstract and again at first use in the main text.

13. Submission

Submit through the MLCS online submission system. Upload:

1. a cover letter;
2. the title page;
3. the anonymized manuscript;
4. the completed reporting-guideline checklist, where applicable; and
5. any supplementary files.

In the cover letter, explain briefly how the work meets the four convergence conditions and disclose any preprint or related publication. You may also suggest up to three potential reviewers who have no conflict of interest with the authors, giving their institutional email addresses, and name any individuals who should not review the manuscript.

14. After submission

• Initial screening. The editors check fit with the scope and the convergence conditions, completeness, compliance with the reporting standards, and textual similarity. A manuscript may be declined at this stage without external review.
• Peer review. A manuscript that passes screening is reviewed double-anonymously by at least two independent reviewers.
• Decision. The editors decide to accept, request minor or major revision, or reject. A revised manuscript must be accompanied by a point-by-point response to every reviewer comment.
• Submissions from editors. A manuscript submitted by a member of the editorial team is handled by an editor with no connection to its authors, and the submitting editor has no access to the review of their own manuscript.

15. After acceptance

• Prepare the final version using the MLCS template.
• Check the proofs promptly. Only corrections of errors are accepted at the proof stage.
• Articles are published open access under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. Authors retain copyright.
• Once an article is published, it can be changed only through a formal correction or retraction notice, in line with COPE guidance.